Most of the research on scheduling and planning in the chemical process industry resorts to general purpose optimisation techniques. When applied to real life industrial problems, these translate into mammoth formulations dominated by the presence of discrete variables, rendering them difficult to solve. This work proposes a fresh viewpoint to approaching these problems by exploiting contextual information to tame the optimisation task. The concept is applied to a delivery scheduling problem for which initial attempts using a pure mathematical programming approach were futile in terms of solving industrial problems. A tailored approach to the problem resulted in a dramatic reduction in the computation time without sacrificing on the quality of the solution. This method combines a tree representation of customer orders and an optimal search criterion to reap benefits of over a quarter of a million dollars per annum for no investment, in a real time application.